Charge transfer at air-water interfaces: A machine learning potential-based molecular dynamics study.

Water at interfaces plays a crucial role in many natural processes and industrial applications. However, the relationship between water's hydrogen bonding and charge transfer characteristics at these interfaces remains poorly understood. Here, we develop machine learning potentials at near density functional theory accuracy based on datasets generated with ab initio molecular dynamics simulations, enabling us to explore the structure and charge transfer at air-water interfaces. Our simulations reveal a non-uniform charge distribution along the interfacial normal direction: water molecules in the outermost layer in direct contact with the air tend to be positively charged, while those in a thin sub-interface layer are negatively charged. We further demonstrate that this uneven charge distribution arises from the donor-acceptor asymmetry of H-bonds among interfacial water molecules. These findings provide a detailed atomic-level insight into the charge transfer behaviors of water at interfaces.